/**
* Check whether the contents of src and dst are the same.
*
* <p>Return false if dstpath does not exist
*
* <p>If the files have different sizes, return false.
*
* <p>If the files have the same sizes, the file checksums will be compared.
*
* <p>When file checksum is not supported in any of file systems, two files are considered as the
* same if they have the same size.
*/
private static boolean sameFile(
FileSystem srcfs, FileStatus srcstatus, FileSystem dstfs, Path dstpath) throws IOException {
FileStatus dststatus;
try {
dststatus = dstfs.getFileStatus(dstpath);
} catch (FileNotFoundException fnfe) {
return false;
}
// same length?
if (srcstatus.getLen() != dststatus.getLen()) {
return false;
}
// compare checksums
try {
final FileChecksum srccs = srcfs.getFileChecksum(srcstatus.getPath());
final FileChecksum dstcs = dstfs.getFileChecksum(dststatus.getPath());
// return true if checksum is not supported
// (i.e. some of the checksums is null)
return srccs == null || dstcs == null || srccs.equals(dstcs);
} catch (FileNotFoundException fnfe) {
return false;
}
}
/**
* Generate the list of files and make them into FileSplits. This needs to be copied to insert a
* filter on acceptable data
*/
@Override
public List<InputSplit> getSplits(JobContext job) throws IOException {
long minSize = Math.max(getFormatMinSplitSize(), getMinSplitSize(job));
long maxSize = getMaxSplitSize(job);
long desiredMappers =
job.getConfiguration().getLong("org.systemsbiology.jxtandem.DesiredXMLInputMappers", 0);
// generate splits
List<InputSplit> splits = new ArrayList<InputSplit>();
List<FileStatus> fileStatuses = listStatus(job);
boolean forceNumberMappers = fileStatuses.size() == 1;
for (FileStatus file : fileStatuses) {
Path path = file.getPath();
if (!isPathAcceptable(path)) // filter acceptable data
continue;
FileSystem fs = path.getFileSystem(job.getConfiguration());
long length = file.getLen();
BlockLocation[] blkLocations = fs.getFileBlockLocations(file, 0, length);
if ((length != 0) && isSplitable(job, path)) {
long blockSize = file.getBlockSize();
// use desired mappers to force more splits
if (forceNumberMappers && desiredMappers > 0)
maxSize = Math.min(maxSize, (length / desiredMappers));
long splitSize = computeSplitSize(blockSize, minSize, maxSize);
long bytesRemaining = length;
while (withinSlop(splitSize, bytesRemaining)) {
int blkIndex = getBlockIndex(blkLocations, length - bytesRemaining);
splits.add(
new FileSplit(
path, length - bytesRemaining, splitSize, blkLocations[blkIndex].getHosts()));
bytesRemaining -= splitSize;
}
if (bytesRemaining != 0) {
splits.add(
new FileSplit(
path,
length - bytesRemaining,
bytesRemaining,
blkLocations[blkLocations.length - 1].getHosts()));
}
} else if (length != 0) {
splits.add(new FileSplit(path, 0, length, blkLocations[0].getHosts()));
} else {
// Create empty hosts array for zero length files
splits.add(new FileSplit(path, 0, length, new String[0]));
}
}
System.out.println("Total # of splits: " + splits.size());
// LOG.debug("Total # of splits: " + splits.size());
return splits;
}
/**
* Combine the status stored in the index and the underlying status.
*
* @param h status stored in the index
* @param cache caching the underlying file statuses
* @return the combined file status
* @throws IOException
*/
private FileStatus toFileStatus(HarStatus h, Map<String, FileStatus> cache) throws IOException {
FileStatus underlying = null;
if (cache != null) {
underlying = cache.get(h.partName);
}
if (underlying == null) {
final Path p = h.isDir ? archivePath : new Path(archivePath, h.partName);
underlying = fs.getFileStatus(p);
if (cache != null) {
cache.put(h.partName, underlying);
}
}
long modTime = 0;
int version = metadata.getVersion();
if (version < 3) {
modTime = underlying.getModificationTime();
} else if (version == 3) {
modTime = h.getModificationTime();
}
return new FileStatus(
h.isDir() ? 0L : h.getLength(),
h.isDir(),
underlying.getReplication(),
underlying.getBlockSize(),
modTime,
underlying.getAccessTime(),
underlying.getPermission(),
underlying.getOwner(),
underlying.getGroup(),
makeRelative(this.uri.getPath(), new Path(h.name)));
}
/**
* Produce splits such that each is no greater than the quotient of the total size and the
* number of splits requested.
*
* @param job The handle to the JobConf object
* @param numSplits Number of splits requested
*/
public InputSplit[] getSplits(JobConf job, int numSplits) throws IOException {
int cnfiles = job.getInt(SRC_COUNT_LABEL, -1);
long cbsize = job.getLong(TOTAL_SIZE_LABEL, -1);
String srcfilelist = job.get(SRC_LIST_LABEL, "");
if (cnfiles < 0 || cbsize < 0 || "".equals(srcfilelist)) {
throw new RuntimeException(
"Invalid metadata: #files("
+ cnfiles
+ ") total_size("
+ cbsize
+ ") listuri("
+ srcfilelist
+ ")");
}
Path src = new Path(srcfilelist);
FileSystem fs = src.getFileSystem(job);
FileStatus srcst = fs.getFileStatus(src);
ArrayList<FileSplit> splits = new ArrayList<FileSplit>(numSplits);
LongWritable key = new LongWritable();
FilePair value = new FilePair();
final long targetsize = cbsize / numSplits;
long pos = 0L;
long last = 0L;
long acc = 0L;
long cbrem = srcst.getLen();
SequenceFile.Reader sl = null;
try {
sl = new SequenceFile.Reader(fs, src, job);
for (; sl.next(key, value); last = sl.getPosition()) {
// if adding this split would put this split past the target size,
// cut the last split and put this next file in the next split.
if (acc + key.get() > targetsize && acc != 0) {
long splitsize = last - pos;
splits.add(new FileSplit(src, pos, splitsize, (String[]) null));
cbrem -= splitsize;
pos = last;
acc = 0L;
}
acc += key.get();
}
} finally {
checkAndClose(sl);
}
if (cbrem != 0) {
splits.add(new FileSplit(src, pos, cbrem, (String[]) null));
}
return splits.toArray(new FileSplit[splits.size()]);
}
// Checks if the cache has already been localized and is fresh
private static boolean ifExistsAndFresh(
Configuration conf,
FileSystem fs,
URI cache,
long confFileStamp,
CacheStatus lcacheStatus,
FileStatus fileStatus)
throws IOException {
// check for existence of the cache
long dfsFileStamp;
if (fileStatus != null) {
dfsFileStamp = fileStatus.getModificationTime();
} else {
dfsFileStamp = getTimestamp(conf, cache);
}
// ensure that the file on hdfs hasn't been modified since the job started
if (dfsFileStamp != confFileStamp) {
LOG.fatal("File: " + cache + " has changed on HDFS since job started");
throw new IOException("File: " + cache + " has changed on HDFS since job started");
}
if (dfsFileStamp != lcacheStatus.mtime) {
// needs refreshing
return false;
}
return true;
}
/** Added for back compatibility. */
public static Path getLocalCache(
URI cache,
Configuration conf,
Path subdir,
FileStatus fileStatus,
boolean isArchive,
long confFileStamp,
Path currentWorkDir,
boolean honorSymLinkConf,
MRAsyncDiskService asyncDiskService,
LocalDirAllocator lDirAllocator)
throws IOException {
return getLocalCache(
cache,
conf,
subdir,
fileStatus,
isArchive,
confFileStamp,
fileStatus.getLen(),
currentWorkDir,
honorSymLinkConf,
asyncDiskService,
lDirAllocator);
}
/**
* Initialize a Har filesystem per har archive. The archive home directory is the top level
* directory in the filesystem that contains the HAR archive. Be careful with this method, you do
* not want to go on creating new Filesystem instances per call to path.getFileSystem(). the uri
* of Har is har://underlyingfsscheme-host:port/archivepath. or har:///archivepath. This assumes
* the underlying filesystem to be used in case not specified.
*/
@Override
public void initialize(URI name, Configuration conf) throws IOException {
// initialize the metadata cache, if needed
initializeMetadataCache(conf);
// decode the name
URI underLyingURI = decodeHarURI(name, conf);
// we got the right har Path- now check if this is
// truly a har filesystem
Path harPath = archivePath(new Path(name.getScheme(), name.getAuthority(), name.getPath()));
if (harPath == null) {
throw new IOException("Invalid path for the Har Filesystem. " + name.toString());
}
if (fs == null) {
fs = FileSystem.get(underLyingURI, conf);
}
uri = harPath.toUri();
archivePath = new Path(uri.getPath());
harAuth = getHarAuth(underLyingURI);
// check for the underlying fs containing
// the index file
Path masterIndexPath = new Path(archivePath, "_masterindex");
Path archiveIndexPath = new Path(archivePath, "_index");
if (!fs.exists(masterIndexPath) || !fs.exists(archiveIndexPath)) {
throw new IOException(
"Invalid path for the Har Filesystem. " + "No index file in " + harPath);
}
metadata = harMetaCache.get(uri);
if (metadata != null) {
FileStatus mStat = fs.getFileStatus(masterIndexPath);
FileStatus aStat = fs.getFileStatus(archiveIndexPath);
if (mStat.getModificationTime() != metadata.getMasterIndexTimestamp()
|| aStat.getModificationTime() != metadata.getArchiveIndexTimestamp()) {
// the archive has been overwritten since we last read it
// remove the entry from the meta data cache
metadata = null;
harMetaCache.remove(uri);
}
}
if (metadata == null) {
metadata = new HarMetaData(fs, masterIndexPath, archiveIndexPath);
metadata.parseMetaData();
harMetaCache.put(uri, metadata);
}
}
private FSDataOutputStream create(Path f, Reporter reporter, FileStatus srcstat)
throws IOException {
if (destFileSys.exists(f)) {
destFileSys.delete(f, false);
}
if (!preserve_status) {
return destFileSys.create(f, true, sizeBuf, reporter);
}
FsPermission permission =
preseved.contains(FileAttribute.PERMISSION) ? srcstat.getPermission() : null;
short replication =
preseved.contains(FileAttribute.REPLICATION)
? srcstat.getReplication()
: destFileSys.getDefaultReplication();
long blockSize =
preseved.contains(FileAttribute.BLOCK_SIZE)
? srcstat.getBlockSize()
: destFileSys.getDefaultBlockSize();
return destFileSys.create(f, permission, true, sizeBuf, replication, blockSize, reporter);
}
/**
* Get block locations from the underlying fs and fix their offsets and lengths.
*
* @param file the input file status to get block locations
* @param start the start of the desired range in the contained file
* @param len the length of the desired range
* @return block locations for this segment of file
* @throws IOException
*/
@Override
public BlockLocation[] getFileBlockLocations(FileStatus file, long start, long len)
throws IOException {
HarStatus hstatus = getFileHarStatus(file.getPath());
Path partPath = new Path(archivePath, hstatus.getPartName());
FileStatus partStatus = metadata.getPartFileStatus(partPath);
// get all part blocks that overlap with the desired file blocks
BlockLocation[] locations =
fs.getFileBlockLocations(partStatus, hstatus.getStartIndex() + start, len);
return fixBlockLocations(locations, start, len, hstatus.getStartIndex());
}
/**
* list subfiles
*
* @param hdfsPath !null path probably exists
* @return !null list of enclused file names - emptu if file of not exists
*/
@Override
public String[] ls(final String hdfsPath) {
if (isRunningAsUser()) {
return super.ls(hdfsPath);
}
final FileSystem fs = getDFS();
try {
FileStatus[] statuses = fs.listStatus(new Path(hdfsPath));
if (statuses == null) return new String[0];
List<String> holder = new ArrayList<String>();
for (int i = 0; i < statuses.length; i++) {
FileStatus statuse = statuses[i];
String s = statuse.getPath().getName();
holder.add(s);
}
String[] ret = new String[holder.size()];
holder.toArray(ret);
return ret;
} catch (IOException e) {
throw new RuntimeException(e);
}
}
public static void recursePath(Configuration conf, Path path, Job job) {
try {
FileSystem fs = path.getFileSystem(conf);
FileStatus[] fstats = fs.listStatus(path);
if (fstats != null) {
for (FileStatus f : fstats) {
Path p = f.getPath();
;
if (fs.isFile(p)) {
// connection times out otherwise
System.err.println("file:" + p.toString());
FileInputFormat.addInputPath(job, p);
} else {
System.err.println("dir:" + p.toString());
recursePath(conf, p, job);
}
}
}
} catch (IOException e) {
// shouldn't be here
throw new RuntimeException(e);
}
}
public void copyInitialState(Path origAppDir) throws IOException {
// locate previous snapshot
String newAppDir = this.dag.assertAppPath();
FSRecoveryHandler recoveryHandler = new FSRecoveryHandler(origAppDir.toString(), conf);
// read snapshot against new dependencies
Object snapshot = recoveryHandler.restore();
if (snapshot == null) {
throw new IllegalArgumentException("No previous application state found in " + origAppDir);
}
InputStream logIs = recoveryHandler.getLog();
// modify snapshot state to switch app id
((StreamingContainerManager.CheckpointState) snapshot).setApplicationId(this.dag, conf);
Path checkpointPath = new Path(newAppDir, LogicalPlan.SUBDIR_CHECKPOINTS);
FileSystem fs = FileSystem.newInstance(origAppDir.toUri(), conf);
// remove the path that was created by the storage agent during deserialization and replacement
fs.delete(checkpointPath, true);
// write snapshot to new location
recoveryHandler = new FSRecoveryHandler(newAppDir, conf);
recoveryHandler.save(snapshot);
OutputStream logOs = recoveryHandler.rotateLog();
IOUtils.copy(logIs, logOs);
logOs.flush();
logOs.close();
logIs.close();
// copy sub directories that are not present in target
FileStatus[] lFiles = fs.listStatus(origAppDir);
for (FileStatus f : lFiles) {
if (f.isDirectory()) {
String targetPath = f.getPath().toString().replace(origAppDir.toString(), newAppDir);
if (!fs.exists(new Path(targetPath))) {
LOG.debug("Copying {} to {}", f.getPath(), targetPath);
FileUtil.copy(fs, f.getPath(), fs, new Path(targetPath), false, conf);
// FSUtil.copy(fs, f, fs, new Path(targetPath), false, false, conf);
} else {
LOG.debug("Ignoring {} as it already exists under {}", f.getPath(), targetPath);
// FSUtil.setPermission(fs, new Path(targetPath), new FsPermission((short)0777));
}
}
}
}
/**
* Get the locally cached file or archive; it could either be previously cached (and valid) or
* copy it from the {@link FileSystem} now.
*
* @param cache the cache to be localized, this should be specified as new
* URI(hdfs://hostname:port/absolute_path_to_file#LINKNAME). If no schema or hostname:port is
* provided the file is assumed to be in the filesystem being used in the Configuration
* @param conf The Confguration file which contains the filesystem
* @param baseDir The base cache Dir where you wnat to localize the files/archives
* @param fileStatus The file status on the dfs.
* @param isArchive if the cache is an archive or a file. In case it is an archive with a .zip or
* .jar or .tar or .tgz or .tar.gz extension it will be unzipped/unjarred/untarred
* automatically and the directory where the archive is unzipped/unjarred/untarred is returned
* as the Path. In case of a file, the path to the file is returned
* @param confFileStamp this is the hdfs file modification timestamp to verify that the file to be
* cached hasn't changed since the job started
* @param currentWorkDir this is the directory where you would want to create symlinks for the
* locally cached files/archives
* @return the path to directory where the archives are unjarred in case of archives, the path to
* the file where the file is copied locally
* @throws IOException
*/
public static Path getLocalCache(
URI cache,
Configuration conf,
Path baseDir,
FileStatus fileStatus,
boolean isArchive,
long confFileStamp,
Path currentWorkDir,
MRAsyncDiskService asyncDiskService)
throws IOException {
return getLocalCache(
cache,
conf,
baseDir,
fileStatus,
isArchive,
confFileStamp,
fileStatus.getLen(),
currentWorkDir,
true,
asyncDiskService,
new LocalDirAllocator("mapred.local.dir"));
}
private static void updatePermissions(
FileStatus src, FileStatus dst, EnumSet<FileAttribute> preseved, FileSystem destFileSys)
throws IOException {
String owner = null;
String group = null;
if (preseved.contains(FileAttribute.USER) && !src.getOwner().equals(dst.getOwner())) {
owner = src.getOwner();
}
if (preseved.contains(FileAttribute.GROUP) && !src.getGroup().equals(dst.getGroup())) {
group = src.getGroup();
}
if (owner != null || group != null) {
destFileSys.setOwner(dst.getPath(), owner, group);
}
if (preseved.contains(FileAttribute.PERMISSION)
&& !src.getPermission().equals(dst.getPermission())) {
destFileSys.setPermission(dst.getPath(), src.getPermission());
}
}
/**
* Copy a file to a destination.
*
* @param srcstat src path and metadata
* @param dstpath dst path
* @param reporter
*/
private void copy(
FileStatus srcstat,
Path relativedst,
OutputCollector<WritableComparable<?>, Text> outc,
Reporter reporter)
throws IOException {
Path absdst = new Path(destPath, relativedst);
int totfiles = job.getInt(SRC_COUNT_LABEL, -1);
assert totfiles >= 0 : "Invalid file count " + totfiles;
// if a directory, ensure created even if empty
if (srcstat.isDir()) {
if (destFileSys.exists(absdst)) {
if (!destFileSys.getFileStatus(absdst).isDir()) {
throw new IOException("Failed to mkdirs: " + absdst + " is a file.");
}
} else if (!destFileSys.mkdirs(absdst)) {
throw new IOException("Failed to mkdirs " + absdst);
}
// TODO: when modification times can be set, directories should be
// emitted to reducers so they might be preserved. Also, mkdirs does
// not currently return an error when the directory already exists;
// if this changes, all directory work might as well be done in reduce
return;
}
if (destFileSys.exists(absdst) && !overwrite && !needsUpdate(srcstat, destFileSys, absdst)) {
outc.collect(null, new Text("SKIP: " + srcstat.getPath()));
++skipcount;
reporter.incrCounter(Counter.SKIP, 1);
updateStatus(reporter);
return;
}
Path tmpfile = new Path(job.get(TMP_DIR_LABEL), relativedst);
long cbcopied = 0L;
FSDataInputStream in = null;
FSDataOutputStream out = null;
try {
// open src file
try {
in = srcstat.getPath().getFileSystem(job).open(srcstat.getPath());
} catch (IOException e) {
LOG.error("Failed to open src file " + srcstat.getPath() + ", ignore and return");
in = null;
return;
}
reporter.incrCounter(Counter.BYTESEXPECTED, srcstat.getLen());
// open tmp file
out = create(tmpfile, reporter, srcstat);
// copy file
for (int cbread; (cbread = in.read(buffer)) >= 0; ) {
out.write(buffer, 0, cbread);
cbcopied += cbread;
reporter.setStatus(
String.format("%.2f ", cbcopied * 100.0 / srcstat.getLen())
+ absdst
+ " [ "
+ StringUtils.humanReadableInt(cbcopied)
+ " / "
+ StringUtils.humanReadableInt(srcstat.getLen())
+ " ]");
}
} finally {
checkAndClose(in);
checkAndClose(out);
}
if (cbcopied != srcstat.getLen()) {
if (srcstat.getLen() == 0 && cbcopied > 0) {
LOG.info("most likely see a WAL file corruption: " + srcstat.getPath());
} else {
throw new IOException(
"File size not matched: copied "
+ bytesString(cbcopied)
+ " to tmpfile (="
+ tmpfile
+ ") but expected "
+ bytesString(srcstat.getLen())
+ " from "
+ srcstat.getPath());
}
} else {
if (totfiles == 1) {
// Copying a single file; use dst path provided by user as destination
// rather than destination directory, if a file
Path dstparent = absdst.getParent();
if (!(destFileSys.exists(dstparent) && destFileSys.getFileStatus(dstparent).isDir())) {
absdst = dstparent;
}
}
if (destFileSys.exists(absdst) && destFileSys.getFileStatus(absdst).isDir()) {
throw new IOException(absdst + " is a directory");
}
if (!destFileSys.mkdirs(absdst.getParent())) {
throw new IOException("Failed to craete parent dir: " + absdst.getParent());
}
rename(tmpfile, absdst);
FileStatus dststat = destFileSys.getFileStatus(absdst);
if (dststat.getLen() != srcstat.getLen()) {
destFileSys.delete(absdst, false);
throw new IOException(
"File size not matched: copied "
+ bytesString(dststat.getLen())
+ " to dst (="
+ absdst
+ ") but expected "
+ bytesString(srcstat.getLen())
+ " from "
+ srcstat.getPath());
}
updatePermissions(srcstat, dststat);
}
// report at least once for each file
++copycount;
reporter.incrCounter(Counter.BYTESCOPIED, cbcopied);
reporter.incrCounter(Counter.COPY, 1);
updateStatus(reporter);
}
void dumpFileStatus(FileStatus fs) {
if (LOG.isDebugEnabled()) {
LOG.debug(
"FileStatus:" + fs.getText() + " " + fs.getColor() + " " + " " + fs.getClass().getName());
}
}
/**
* Return true if dst should be replaced by src and the update flag is set. Right now, this
* merely checks that the src and dst len are not equal. This should be improved on once
* modification times, CRCs, etc. can be meaningful in this context.
*
* @throws IOException
*/
private boolean needsUpdate(FileStatus srcstatus, FileSystem dstfs, Path dstpath)
throws IOException {
return update && !sameFile(srcstatus.getPath().getFileSystem(job), srcstatus, dstfs, dstpath);
}
// Information needed to get a single file:
// BASE_PATH, FILE_ID, TIMESTAMP_START, TIMESTAMP_STOP, SOURCE, FILESYSTEM
private static Vector<Path> getFile(
FileSystem fs, Hashtable<String, String> config, dbutil db_util) throws Exception {
Long latestVersion = latestVersion(config, db_util);
try {
config.put("timestamp_start", config.get("timestamp_start"));
config.put("timestamp_real", latestVersion.toString());
config.put("timestamp_stop", latestVersion.toString());
} catch (Exception E) {
logger.error("Tryign to get file that is impossible to generate: " + getFullPath(config));
return null;
}
if (Integer.parseInt(config.get("timestamp_start"))
> Integer.parseInt(config.get("timestamp_stop"))) {
return null;
}
logger.debug(
"Getting DB for timestamp "
+ config.get("timestamp_start")
+ " to "
+ config.get("timestamp_stop"));
String final_result = getFullPath(config);
String temp_path_base =
config.get("local_temp_path")
+ "_"
+ config.get("task_id")
+ "_"
+ config.get("run_id")
+ "/";
Path newPath = new Path(final_result + "*");
Vector<Path> ret_path = new Vector<Path>();
String lockName = lock(final_result.replaceAll("/", "_"));
if (fs.globStatus(newPath).length != 0) {
ret_path.add(newPath);
unlock(lockName);
config.put("full_file_name", final_result);
return ret_path;
} else {
if (!config.get("source").equals("local")) {
config.put("temp_path_base", temp_path_base);
config.put("timestamp_start", config.get("timestamp_start"));
config.put("timestamp_real", latestVersion.toString());
config.put("timestamp_stop", latestVersion.toString());
Class<?> sourceClass =
Class.forName("org.gestore.plugin.source." + config.get("source") + "Source");
Method process_data = sourceClass.getMethod("process", Hashtable.class, FileSystem.class);
Object processor = sourceClass.newInstance();
Object retVal;
try {
retVal = process_data.invoke(processor, config, fs);
} catch (InvocationTargetException E) {
Throwable exception = E.getTargetException();
logger.error("Unable to call method in child class: " + exception.toString());
exception.printStackTrace(System.out);
unlock(lockName);
return null;
}
FileStatus[] files = (FileStatus[]) retVal;
if (files == null) {
logger.error("Error getting files, no files returned");
return null;
}
for (FileStatus file : files) {
Path cur_file = file.getPath();
Path cur_local_path = new Path(temp_path_base + config.get("file_id"));
String suffix = getSuffix(config.get("file_id"), cur_file.getName());
cur_local_path = cur_local_path.suffix(suffix);
Path res_path = new Path(new String(final_result + suffix));
logger.debug("Moving file" + cur_file.toString() + " to " + res_path.toString());
if (config.get("copy").equals("true")) {
fs.moveFromLocalFile(cur_file, res_path);
} else {
fs.rename(cur_file, res_path);
}
}
config.put("full_file_name", final_result);
}
}
unlock(lockName);
return ret_path;
}
/** * Compare the checksums of the hdfs file as well as the local copied file. * * @author [email protected] * @date Fri Jan 27 06:06:00 2012 */ boolean compareChecksums(FileSystem fs, Path p, String sFsPath) { try { // get hdfs file info FileStatus stat = fs.getFileStatus(p); // get HDFS checksum FileChecksum ck = fs.getFileChecksum(p); String sCk, sCkShort; if (ck == null) { sCk = sCkShort = "<null>"; } else { sCk = ck.toString(); sCkShort = sCk.replaceAll("^.*:", ""); } // System.out.println(p.toUri().getPath() + " len=" + stat.getLen() // + " " + stat.getOwner() + "/" + stat.getGroup() // + " checksum=" + sCk); // find the local file File fLocal = new File(sFsPath); if (!fLocal.exists()) { System.out.println("CHECKSUM-ERROR: file does not exist: " + sFsPath); return false; } if (!fLocal.isFile()) { System.out.println("CHECKSUM-ERROR: path is not a file: " + sFsPath); return false; } if (stat.getLen() != fLocal.length()) { System.out.println( "CHECKSUM-ERROR: length mismatch: " + sFsPath + " hdfslen=" + stat.getLen() + " fslen=" + fLocal.length()); return false; } // get local fs checksum FileChecksum ckLocal = getLocalFileChecksum(sFsPath); if (ckLocal == null) { System.out.println("ERROR Failed to get checksum for local file " + sFsPath); return false; } // compare checksums as a string, after stripping the // algorithm name from the beginning String sCkLocal = ckLocal.toString(); String sCkLocalShort = sCkLocal.replaceAll("^.*:", ""); if (false == sCkShort.equals(sCkLocalShort)) { System.out.println( "CHECKSUM-ERROR: checksum mismatch: " + sFsPath + "\nhdfs = " + sCk + "\nlocal= " + sCkLocal); return false; } return true; } catch (IOException e) { System.out.println("CHECKSUM-ERROR: " + sFsPath + " exception " + e.toString()); } return false; }
/** Delete the dst files/dirs which do not exist in src */
private static void deleteNonexisting(
FileSystem dstfs,
FileStatus dstroot,
Path dstsorted,
FileSystem jobfs,
Path jobdir,
JobConf jobconf,
Configuration conf)
throws IOException {
if (!dstroot.isDir()) {
throw new IOException(
"dst must be a directory when option "
+ Options.DELETE.cmd
+ " is set, but dst (= "
+ dstroot.getPath()
+ ") is not a directory.");
}
// write dst lsr results
final Path dstlsr = new Path(jobdir, "_distcp_dst_lsr");
final SequenceFile.Writer writer =
SequenceFile.createWriter(
jobfs,
jobconf,
dstlsr,
Text.class,
FileStatus.class,
SequenceFile.CompressionType.NONE);
try {
// do lsr to get all file statuses in dstroot
final Stack<FileStatus> lsrstack = new Stack<FileStatus>();
for (lsrstack.push(dstroot); !lsrstack.isEmpty(); ) {
final FileStatus status = lsrstack.pop();
if (status.isDir()) {
for (FileStatus child : dstfs.listStatus(status.getPath())) {
String relative = makeRelative(dstroot.getPath(), child.getPath());
writer.append(new Text(relative), child);
lsrstack.push(child);
}
}
}
} finally {
checkAndClose(writer);
}
// sort lsr results
final Path sortedlsr = new Path(jobdir, "_distcp_dst_lsr_sorted");
SequenceFile.Sorter sorter =
new SequenceFile.Sorter(
jobfs, new Text.Comparator(), Text.class, FileStatus.class, jobconf);
sorter.sort(dstlsr, sortedlsr);
// compare lsr list and dst list
SequenceFile.Reader lsrin = null;
SequenceFile.Reader dstin = null;
try {
lsrin = new SequenceFile.Reader(jobfs, sortedlsr, jobconf);
dstin = new SequenceFile.Reader(jobfs, dstsorted, jobconf);
// compare sorted lsr list and sorted dst list
final Text lsrpath = new Text();
final FileStatus lsrstatus = new FileStatus();
final Text dstpath = new Text();
final Text dstfrom = new Text();
final FsShell shell = new FsShell(conf);
final String[] shellargs = {"-rmr", null};
boolean hasnext = dstin.next(dstpath, dstfrom);
for (; lsrin.next(lsrpath, lsrstatus); ) {
int dst_cmp_lsr = dstpath.compareTo(lsrpath);
for (; hasnext && dst_cmp_lsr < 0; ) {
hasnext = dstin.next(dstpath, dstfrom);
dst_cmp_lsr = dstpath.compareTo(lsrpath);
}
if (dst_cmp_lsr == 0) {
// lsrpath exists in dst, skip it
hasnext = dstin.next(dstpath, dstfrom);
} else {
// lsrpath does not exist, delete it
String s = new Path(dstroot.getPath(), lsrpath.toString()).toString();
if (shellargs[1] == null || !isAncestorPath(shellargs[1], s)) {
shellargs[1] = s;
int r = 0;
try {
r = shell.run(shellargs);
} catch (Exception e) {
throw new IOException("Exception from shell.", e);
}
if (r != 0) {
throw new IOException(
"\"" + shellargs[0] + " " + shellargs[1] + "\" returns non-zero value " + r);
}
}
}
}
} finally {
checkAndClose(lsrin);
checkAndClose(dstin);
}
}
public int run(String[] args) throws Exception {
// printUsage();
/*
* SETUP
*/
Configuration argConf = getConf();
Hashtable<String, String> confArg = new Hashtable<String, String>();
setup(confArg, argConf);
Date currentTime = new Date();
Date endDate = new Date(new Long(confArg.get("timestamp_stop")));
Boolean full_run = confArg.get("intermediate").matches("(?i).*true.*");
Boolean quick_add = confArg.get("quick_add").matches("(?i).*true.*");
logger.info("Running GeStore");
// ZooKeeper setup
Configuration config = HBaseConfiguration.create();
zkWatcher = new ZooKeeperWatcher(config, "Testing", new HBaseAdmin(config));
zkInstance =
new ZooKeeper(
ZKConfig.getZKQuorumServersString(config),
config.getInt("zookeeper.session.timeout", -1),
zkWatcher);
if (!confArg.get("task_id").isEmpty()) {
confArg.put("temp_path", confArg.get("temp_path") + confArg.get("task_id"));
}
String lockRequest = confArg.get("file_id");
if (!confArg.get("run_id").isEmpty())
lockRequest = lockRequest + "_" + confArg.get("run_id") + "_";
if (!confArg.get("task_id").isEmpty())
lockRequest = lockRequest + "_" + confArg.get("task_id") + "_";
// Get type of movement
toFrom type_move = checkArgs(confArg);
if (type_move == toFrom.LOCAL2REMOTE && !confArg.get("format").equals("unknown")) {
List<String> arguments = new ArrayList<String>();
arguments.add("-Dinput=" + confArg.get("local_path"));
arguments.add("-Dtable=" + confArg.get("file_id"));
arguments.add("-Dtimestamp=" + confArg.get("timestamp_stop"));
arguments.add("-Dtype=" + confArg.get("format"));
arguments.add("-Dtarget_dir=" + confArg.get("base_path") + "_" + confArg.get("file_id"));
arguments.add("-Dtemp_hdfs_path=" + confArg.get("temp_path"));
arguments.add("-Drun_id=" + confArg.get("run_id"));
if (!confArg.get("run_id").isEmpty()) arguments.add("-Drun_id=" + confArg.get("run_id"));
if (!confArg.get("task_id").isEmpty()) arguments.add("-Dtask_id=" + confArg.get("task_id"));
if (quick_add) arguments.add("-Dquick_add=" + confArg.get("quick_add"));
String lockName = lock(lockRequest);
String[] argumentString = arguments.toArray(new String[arguments.size()]);
adddb.main(argumentString);
unlock(lockName);
System.exit(0);
}
// Database registration
dbutil db_util = new dbutil(config);
db_util.register_database(confArg.get("db_name_files"), true);
db_util.register_database(confArg.get("db_name_runs"), true);
db_util.register_database(confArg.get("db_name_updates"), true);
FileSystem hdfs = FileSystem.get(config);
FileSystem localFS = FileSystem.getLocal(config);
// Get source type
confArg.put("source", getSource(db_util, confArg.get("db_name_files"), confArg.get("file_id")));
confArg.put(
"database", isDatabase(db_util, confArg.get("db_name_files"), confArg.get("file_id")));
if (!confArg.get("source").equals("local")
&& type_move == toFrom.REMOTE2LOCAL
&& !confArg.get("timestamp_stop").equals(Integer.toString(Integer.MAX_VALUE))) {
confArg.put("timestamp_stop", Long.toString(latestVersion(confArg, db_util)));
}
/*
* Get previous timestamp
*/
Get run_id_get = new Get(confArg.get("run_id").getBytes());
Result run_get = db_util.doGet(confArg.get("db_name_runs"), run_id_get);
KeyValue run_file_prev =
run_get.getColumnLatest(
"d".getBytes(), (confArg.get("file_id") + "_db_timestamp").getBytes());
String last_timestamp = new String("0");
if (null != run_file_prev && !confArg.get("source").equals("local")) {
long last_timestamp_real = run_file_prev.getTimestamp();
Long current_timestamp = new Long(confArg.get("timestamp_real"));
if ((current_timestamp - last_timestamp_real) > 36000) {
last_timestamp = new String(run_file_prev.getValue());
Integer lastTimestamp = new Integer(last_timestamp);
lastTimestamp += 1;
last_timestamp = lastTimestamp.toString();
logger.info("Last timestamp: " + last_timestamp + " End data: " + endDate);
Date last_run = new Date(run_file_prev.getTimestamp());
if (last_run.before(endDate) && !full_run) {
confArg.put("timestamp_start", last_timestamp);
}
}
}
Integer tse = new Integer(confArg.get("timestamp_stop"));
Integer tss = new Integer(confArg.get("timestamp_start"));
if (tss > tse) {
logger.info("No new version of requested file.");
return 0;
}
/*
* Generate file
*/
String lockName = lock(lockRequest);
Get file_id_get = new Get(confArg.get("file_id").getBytes());
Result file_get = db_util.doGet(confArg.get("db_name_files"), file_id_get);
if (!file_get.isEmpty()) {
boolean found =
hasFile(
db_util,
hdfs,
confArg.get("db_name_files"),
confArg.get("file_id"),
getFullPath(confArg));
if (confArg.get("source").equals("fullfile")) {
found = false;
}
String filenames_put =
getFileNames(
db_util, confArg.get("db_name_files"), confArg.get("file_id"), getFullPath(confArg));
// Filename not found in file database
if (!found && type_move == toFrom.REMOTE2LOCAL) {
if (!confArg.get("source").equals("local")) {
// Generate intermediate file
if (getFile(hdfs, confArg, db_util) == null) {
unlock(lockName);
return 1;
}
// Put generated file into file database
if (!confArg.get("format").equals("fullfile")) {
putFileEntry(
db_util,
hdfs,
confArg.get("db_name_files"),
confArg.get("file_id"),
confArg.get("full_file_name"),
confArg.get("source"));
}
} else {
logger.warn("Remote file not found, and cannot be generated! File: " + confArg);
unlock(lockName);
return 1;
}
}
} else {
if (type_move == toFrom.REMOTE2LOCAL) {
logger.warn("Remote file not found, and cannot be generated.");
unlock(lockName);
return 1;
}
}
/*
* Copy file
* Update tables
*/
if (type_move == toFrom.LOCAL2REMOTE) {
if (!confArg.get("format").equals("fullfile")) {
putFileEntry(
db_util,
hdfs,
confArg.get("db_name_files"),
confArg.get("file_id"),
getFullPath(confArg),
confArg.get("source"));
}
putRunEntry(
db_util,
confArg.get("db_name_runs"),
confArg.get("run_id"),
confArg.get("file_id"),
confArg.get("type"),
confArg.get("timestamp_real"),
confArg.get("timestamp_stop"),
getFullPath(confArg),
confArg.get("delimiter"));
hdfs.copyFromLocalFile(new Path(confArg.get("local_path")), new Path(getFullPath(confArg)));
} else if (type_move == toFrom.REMOTE2LOCAL) {
FileStatus[] files = hdfs.globStatus(new Path(getFullPath(confArg) + "*"));
putRunEntry(
db_util,
confArg.get("db_name_runs"),
confArg.get("run_id"),
confArg.get("file_id"),
confArg.get("type"),
confArg.get("timestamp_real"),
confArg.get("timestamp_stop"),
getFullPath(confArg),
confArg.get("delimiter"));
unlock(lockName);
for (FileStatus file : files) {
Path cur_file = file.getPath();
Path cur_local_path =
new Path(new String(confArg.get("local_path") + confArg.get("file_id")));
String suffix = getSuffix(getFileName(confArg), cur_file.getName());
if (suffix.length() > 0) {
cur_local_path = cur_local_path.suffix(new String("." + suffix));
}
if (confArg.get("copy").equals("true")) {
String crc = hdfs.getFileChecksum(cur_file).toString();
if (checksumLocalTest(cur_local_path, crc)) {
continue;
} else {
hdfs.copyToLocalFile(cur_file, cur_local_path);
writeChecksum(cur_local_path, crc);
}
} else {
System.out.println(cur_local_path + "\t" + cur_file);
}
}
}
unlock(lockName);
return 0;
}
/**
* Method to move files from HDFS to local filesystem
*
* <p>localPath: Path on the machines filesystem fs:FileSystem object from HDFS pathList:List of
* paths for files that might need to be backed up size:max size in bytes to be backed up
*
* <p>ReturnsDate of the last files backed up if reached size limit, else, zero
*/
public long backupFiles(
String localPath, String preservePath, FileSystem fs, ArrayList<Path> pathList, long size) {
Path fsPath;
long tmpSize = 0;
long tmpDate = 0;
// Start iterating over all paths
for (Path hdfsPath : pathList) {
try {
long nFileSize = fs.getContentSummary(hdfsPath).getLength();
tmpSize = tmpSize + nFileSize;
if ((tmpSize <= size) || (size == 0)) {
FileStatus stat = fs.getFileStatus(hdfsPath);
System.err.println(
"File "
+ hdfsPath.toUri().getPath()
+ " "
+ nFileSize
+ " bytes, "
+ "perms: "
+ stat.getOwner()
+ "/"
+ stat.getGroup()
+ ", "
+ stat.getPermission().toString());
tmpDate = stat.getModificationTime() / 1000;
String sFsPath = localPath + hdfsPath.toUri().getPath();
fsPath = new Path(sFsPath);
File f = new File(sFsPath);
// COMMENTED OUT: until a few backup cycles run
// and the mtime gets in fact set on all copied
// files.
//
// ignore it if the file exists and has the same mtime
// if (f.exists() && f.isFile() && f.lastModified() == stat.getModificationTime())
// {
// System.out.println("no need to backup " + f.toString() + ", mtime matches hdfs");
// continue;
// }
if (false == m_bDryRun) {
// check if we need to back up the local file
// (not directory), if it already exists.
if (f.exists() && f.isFile()) {
// ignore files with substrings in the
// no-preserve file
if (true == doPreserveFile(sFsPath)) {
// move it to the backup path
String sNewPath = preservePath + hdfsPath.toUri().getPath();
File newFile = new File(sNewPath);
// create directory structure for new file?
if (false == newFile.getParentFile().exists()) {
if (false == newFile.getParentFile().mkdirs()) {
System.err.println("Failed to mkdirs " + newFile.getParentFile().toString());
System.exit(1);
}
}
// rename existing file to new location
if (false == f.renameTo(newFile)) {
System.err.println(
"Failed to renameTo " + f.toString() + " to " + newFile.toString());
System.exit(1);
}
System.out.println("preserved " + f.toString() + " into " + newFile.toString());
} else {
System.out.println("skipped preservation of " + f.toString());
}
}
// copy from hdfs to local filesystem
fs.copyToLocalFile(hdfsPath, fsPath);
// set the mtime to match hdfs file
f.setLastModified(stat.getModificationTime());
// compare checksums on both files
compareChecksums(fs, hdfsPath, sFsPath);
}
// don't print the progress after every file -- go
// by at least 1% increments
long nPercentDone = (long) (100 * tmpSize / m_nTotalBytes);
if (nPercentDone > m_nLastPercentBytesDone) {
System.out.println(
"progress: copied "
+ prettyPrintBytes(tmpSize)
+ ", "
+ nPercentDone
+ "% done"
+ ", tstamp="
+ tmpDate);
m_nLastPercentBytesDone = nPercentDone;
}
if (m_nSleepSeconds > 0) {
try {
Thread.sleep(1000 * m_nSleepSeconds);
} catch (Exception e2) {
// ignore
}
}
} else {
return tmpDate;
}
} catch (IOException e) {
System.err.println("FATAL ERROR: Something wrong with the file");
System.err.println(e);
System.out.println(tmpDate);
System.exit(1);
return 0;
}
}
return 0;
}
private void parseMetaData() throws IOException {
Text line = new Text();
long read;
FSDataInputStream in = null;
LineReader lin = null;
try {
in = fs.open(masterIndexPath);
FileStatus masterStat = fs.getFileStatus(masterIndexPath);
masterIndexTimestamp = masterStat.getModificationTime();
lin = new LineReader(in, getConf());
read = lin.readLine(line);
// the first line contains the version of the index file
String versionLine = line.toString();
String[] arr = versionLine.split(" ");
version = Integer.parseInt(arr[0]);
// make it always backwards-compatible
if (this.version > HarFileSystem.VERSION) {
throw new IOException(
"Invalid version " + this.version + " expected " + HarFileSystem.VERSION);
}
// each line contains a hashcode range and the index file name
String[] readStr;
while (read < masterStat.getLen()) {
int b = lin.readLine(line);
read += b;
readStr = line.toString().split(" ");
int startHash = Integer.parseInt(readStr[0]);
int endHash = Integer.parseInt(readStr[1]);
stores.add(
new Store(
Long.parseLong(readStr[2]), Long.parseLong(readStr[3]), startHash, endHash));
line.clear();
}
} catch (IOException ioe) {
LOG.warn("Encountered exception ", ioe);
throw ioe;
} finally {
IOUtils.cleanup(LOG, lin, in);
}
FSDataInputStream aIn = fs.open(archiveIndexPath);
try {
FileStatus archiveStat = fs.getFileStatus(archiveIndexPath);
archiveIndexTimestamp = archiveStat.getModificationTime();
LineReader aLin;
// now start reading the real index file
for (Store s : stores) {
read = 0;
aIn.seek(s.begin);
aLin = new LineReader(aIn, getConf());
while (read + s.begin < s.end) {
int tmp = aLin.readLine(line);
read += tmp;
String lineFeed = line.toString();
String[] parsed = lineFeed.split(" ");
parsed[0] = decodeFileName(parsed[0]);
archive.put(new Path(parsed[0]), new HarStatus(lineFeed));
line.clear();
}
}
} finally {
IOUtils.cleanup(LOG, aIn);
}
}
/**
* Method to go though the HDFS filesystem in a DFS to find all files
*
* <p>fs:FileSystem object from HDFS minDate: Oldest date for files to be backed up maxDate:Newest
* date for files to be backed up p:Path in HDFS to look for files pathList:Will be filled with
* all files in p hmTimestamps: hashmap of timestamps for later sorting
*/
public void checkDir(
FileSystem fs,
long minDate,
long maxDate,
Path p,
ArrayList<Path> pathList,
HashMap<Path, Long> hmTimestamps) {
long tmpDate;
FileStatus[] fStat;
try {
String sPath = p.toUri().getPath();
// If this is a directory
if (fs.getFileStatus(p).isDir()) {
// ignore certain directories
if ("dfstmp".equals(p.getName())
|| "tmp".equals(p.getName())
|| "jobtracker".equals(p.getName())
|| sPath.startsWith("/mapred")
|| "ops".equals(p.getName())
|| p.getName().startsWith("_distcp_logs")) {
return;
}
// dump the mkdir and chmod commands for this
// directory -- skip root directory only
{
FileStatus stat = fs.getFileStatus(p);
if (!sPath.equals("/")) {
m_wrMkdirs.println("hadoop fs -mkdir " + sPath);
}
m_wrChmods.println(
"hadoop fs -chown " + stat.getOwner() + ":" + stat.getGroup() + " " + sPath);
Short sh = new Short(stat.getPermission().toShort());
m_wrChmods.println(
"hadoop fs -chmod " + Long.toOctalString(sh.longValue()) + " " + sPath);
}
fStat = fs.listStatus(p);
// Do a recursive call to all elements
for (int i = 0; i < fStat.length; i++) {
checkDir(fs, minDate, maxDate, fStat[i].getPath(), pathList, hmTimestamps);
}
} else {
// If not a directory then we've found a file
// ignore crc files
if (p.getName().endsWith(".crc")) {
return;
}
// ignore other files
if (sPath.startsWith("/user/oozie/etl/workflows/")) {
return;
}
// try to get the table name from the path. There are
// various types of tables, from those replicated from
// another database to regular hive tables to
// partitioned hive tables. We use table names to
// both exclude some from the backup, and for the rest
// to dump out the schema and partition name.
if (m_ignoreTables != null && m_ignoreTables.doIgnoreFile(sPath)) {
m_nIgnoredTables++;
if (m_nIgnoredTables < 5) {
System.out.println("Skipping ignore-table file: " + sPath);
} else if (m_nIgnoredTables == 5) {
System.out.println("(...not showing other skipped tables...)");
}
return;
}
FileStatus stat = fs.getFileStatus(p);
tmpDate = stat.getModificationTime() / 1000;
// store the chmods/chowns for all files
m_wrChmods.println(
"hadoop fs -chown " + stat.getOwner() + ":" + stat.getGroup() + " " + sPath);
m_wrChmods.println("hadoop fs -chmod " + stat.getPermission().toShort() + " " + sPath);
// check dates. is it too young?
if (tmpDate < minDate) {
return;
}
// is the file too recent?
if (tmpDate > maxDate) {
// System.out.println("file too recent: " + sPath);
return;
}
// file timestamp is ok
pathList.add(p);
hmTimestamps.put(p, new Long(tmpDate));
// store info about total bytes neeed to backup
m_nTotalBytes += fs.getContentSummary(p).getLength();
}
} catch (IOException e) {
System.err.println("ERROR: could not open " + p + ": " + e);
// System.exit(1) ;
}
}
/**
* Initialize DFSCopyFileMapper specific job-configuration.
*
* @param conf : The dfs/mapred configuration.
* @param jobConf : The handle to the jobConf object to be initialized.
* @param args Arguments
*/
private static void setup(Configuration conf, JobConf jobConf, final Arguments args)
throws IOException {
jobConf.set(DST_DIR_LABEL, args.dst.toUri().toString());
// set boolean values
final boolean update = args.flags.contains(Options.UPDATE);
final boolean overwrite = !update && args.flags.contains(Options.OVERWRITE);
jobConf.setBoolean(Options.UPDATE.propertyname, update);
jobConf.setBoolean(Options.OVERWRITE.propertyname, overwrite);
jobConf.setBoolean(
Options.IGNORE_READ_FAILURES.propertyname,
args.flags.contains(Options.IGNORE_READ_FAILURES));
jobConf.setBoolean(
Options.PRESERVE_STATUS.propertyname, args.flags.contains(Options.PRESERVE_STATUS));
final String randomId = getRandomId();
JobClient jClient = new JobClient(jobConf);
Path jobDirectory = new Path(jClient.getSystemDir(), NAME + "_" + randomId);
jobConf.set(JOB_DIR_LABEL, jobDirectory.toString());
FileSystem dstfs = args.dst.getFileSystem(conf);
boolean dstExists = dstfs.exists(args.dst);
boolean dstIsDir = false;
if (dstExists) {
dstIsDir = dstfs.getFileStatus(args.dst).isDir();
}
// default logPath
Path logPath = args.log;
if (logPath == null) {
String filename = "_distcp_logs_" + randomId;
if (!dstExists || !dstIsDir) {
Path parent = args.dst.getParent();
if (!dstfs.exists(parent)) {
dstfs.mkdirs(parent);
}
logPath = new Path(parent, filename);
} else {
logPath = new Path(args.dst, filename);
}
}
FileOutputFormat.setOutputPath(jobConf, logPath);
// create src list, dst list
FileSystem jobfs = jobDirectory.getFileSystem(jobConf);
Path srcfilelist = new Path(jobDirectory, "_distcp_src_files");
jobConf.set(SRC_LIST_LABEL, srcfilelist.toString());
SequenceFile.Writer src_writer =
SequenceFile.createWriter(
jobfs,
jobConf,
srcfilelist,
LongWritable.class,
FilePair.class,
SequenceFile.CompressionType.NONE);
Path dstfilelist = new Path(jobDirectory, "_distcp_dst_files");
SequenceFile.Writer dst_writer =
SequenceFile.createWriter(
jobfs, jobConf, dstfilelist, Text.class, Text.class, SequenceFile.CompressionType.NONE);
Path dstdirlist = new Path(jobDirectory, "_distcp_dst_dirs");
jobConf.set(DST_DIR_LIST_LABEL, dstdirlist.toString());
SequenceFile.Writer dir_writer =
SequenceFile.createWriter(
jobfs,
jobConf,
dstdirlist,
Text.class,
FilePair.class,
SequenceFile.CompressionType.NONE);
// handle the case where the destination directory doesn't exist
// and we've only a single src directory OR we're updating/overwriting
// the contents of the destination directory.
final boolean special = (args.srcs.size() == 1 && !dstExists) || update || overwrite;
int srcCount = 0, cnsyncf = 0, dirsyn = 0;
long fileCount = 0L, byteCount = 0L, cbsyncs = 0L;
try {
for (Iterator<Path> srcItr = args.srcs.iterator(); srcItr.hasNext(); ) {
final Path src = srcItr.next();
FileSystem srcfs = src.getFileSystem(conf);
FileStatus srcfilestat = srcfs.getFileStatus(src);
Path root = special && srcfilestat.isDir() ? src : src.getParent();
if (srcfilestat.isDir()) {
++srcCount;
}
Stack<FileStatus> pathstack = new Stack<FileStatus>();
for (pathstack.push(srcfilestat); !pathstack.empty(); ) {
FileStatus cur = pathstack.pop();
FileStatus[] children = srcfs.listStatus(cur.getPath());
for (int i = 0; i < children.length; i++) {
boolean skipfile = false;
final FileStatus child = children[i];
final String dst = makeRelative(root, child.getPath());
++srcCount;
if (child.isDir()) {
pathstack.push(child);
} else {
// skip file if the src and the dst files are the same.
skipfile = update && sameFile(srcfs, child, dstfs, new Path(args.dst, dst));
// skip file if it exceed file limit or size limit
skipfile |=
fileCount == args.filelimit || byteCount + child.getLen() > args.sizelimit;
if (!skipfile) {
++fileCount;
byteCount += child.getLen();
if (LOG.isTraceEnabled()) {
LOG.trace("adding file " + child.getPath());
}
++cnsyncf;
cbsyncs += child.getLen();
if (cnsyncf > SYNC_FILE_MAX || cbsyncs > BYTES_PER_MAP) {
src_writer.sync();
dst_writer.sync();
cnsyncf = 0;
cbsyncs = 0L;
}
}
}
if (!skipfile) {
src_writer.append(
new LongWritable(child.isDir() ? 0 : child.getLen()), new FilePair(child, dst));
}
dst_writer.append(new Text(dst), new Text(child.getPath().toString()));
}
if (cur.isDir()) {
String dst = makeRelative(root, cur.getPath());
dir_writer.append(new Text(dst), new FilePair(cur, dst));
if (++dirsyn > SYNC_FILE_MAX) {
dirsyn = 0;
dir_writer.sync();
}
}
}
}
} finally {
checkAndClose(src_writer);
checkAndClose(dst_writer);
checkAndClose(dir_writer);
}
FileStatus dststatus = null;
try {
dststatus = dstfs.getFileStatus(args.dst);
} catch (FileNotFoundException fnfe) {
LOG.info(args.dst + " does not exist.");
}
// create dest path dir if copying > 1 file
if (dststatus == null) {
if (srcCount > 1 && !dstfs.mkdirs(args.dst)) {
throw new IOException("Failed to create" + args.dst);
}
}
final Path sorted = new Path(jobDirectory, "_distcp_sorted");
checkDuplication(jobfs, dstfilelist, sorted, conf);
if (dststatus != null && args.flags.contains(Options.DELETE)) {
deleteNonexisting(dstfs, dststatus, sorted, jobfs, jobDirectory, jobConf, conf);
}
Path tmpDir =
new Path(
(dstExists && !dstIsDir) || (!dstExists && srcCount == 1)
? args.dst.getParent()
: args.dst,
"_distcp_tmp_" + randomId);
jobConf.set(TMP_DIR_LABEL, tmpDir.toUri().toString());
LOG.info("srcCount=" + srcCount);
jobConf.setInt(SRC_COUNT_LABEL, srcCount);
jobConf.setLong(TOTAL_SIZE_LABEL, byteCount);
setMapCount(byteCount, jobConf);
}
public int run(String[] args) throws IOException, ClassNotFoundException, InterruptedException {
if (args.length != 2) {
System.out.println("Usage: FeatureMatching ID <inputName.jpeg/inputName.jpg>");
System.exit(1);
}
SimpleDateFormat sdf = new SimpleDateFormat("", Locale.US);
sdf.applyPattern("yyyy-MM-dd_HH-mm-ss");
String time = sdf.format(new Date());
Job job = Job.getInstance();
ID = "/" + args[0];
String filename = args[1];
filename = filename.toLowerCase();
System.out.println("current filename:" + filename);
// Detect illegal username (if the username dir doesn't exist)
File userPath = new File(LOCAL_USER_DIR + ID);
if (!userPath.exists()) {
System.out.println("Unauthorized username!!!\nExiting......");
System.exit(1);
}
// Preprocess the input image.jpg from local dir: /local.../user/ID/input/image.jpg
// Save the features to local dir: hdfs://.../user/ID/input/image.jpg
extractQueryFeatures2HDFS(filename, job);
// Add the feature file to the hdfs cache
String featureFileName = filename.substring(0, filename.lastIndexOf(".")) + ".json";
// job.addCacheFile(new Path(HDFS_HOME + USER + ID + INPUT + "/" +
// featureFileName).toUri());
job.getConfiguration()
.set("featureFilePath", HDFS_HOME + USER + ID + INPUT + "/" + featureFileName);
// Check the file type. Only support jpeg/jpg type images
String type = filename.substring(args[1].lastIndexOf("."));
if (!(type.equals(".jpg") || type.equals(".jpeg"))) {
System.out.println("Image type not supported!!!\nExiting");
System.exit(1);
}
// Input: hdfs://.../features/
// The feature dir is a location of all features extracted from the database
String inputPathStr = HDFS_HOME + FEATURES;
// Output: hdfs://.../user/ID/output/
String outputPathStr = HDFS_HOME + USER + ID + OUTPUT + "/" + time;
job.setInputFormatClass(KeyValueTextInputFormat.class);
// job.setOutputFormatClass(TextOutputFormat.class);
// Get the lists of all feature files: /.../features/data/part-*
FileSystem fs = FileSystem.get(job.getConfiguration());
FileStatus[] statuses = fs.listStatus(new Path(inputPathStr));
StringBuffer sb = new StringBuffer();
for (FileStatus fileStatus : statuses) {
sb.append(fileStatus.getPath() + ",");
}
sb.deleteCharAt(sb.lastIndexOf(","));
job.setJarByClass(FeatureMatching.class);
job.setMapperClass(FeatureMatchMapper.class);
job.setReducerClass(FeatureMatchReducer.class);
// only need one reducer to collect the result
job.setNumReduceTasks(1);
job.setMapOutputKeyClass(IntWritable.class);
job.setMapOutputValueClass(Text.class);
job.setOutputKeyClass(Text.class);
job.setOutputValueClass(DoubleWritable.class);
// Input a directory, so need the recursive input
FileInputFormat.setInputDirRecursive(job, true);
// Set the PathFilter, to omit _SUCCESS files
// (This is not working correctly, as the PathFilter class is an interface rather than a class.
// But the 2nd arg asks me to extend the PathFilter)
// FileInputFormat.setInputPathFilter(job, MyPathFilter.class);
//
// FileInputFormat.setInputPaths(job, new Path(inputPathStr));
FileInputFormat.setInputPaths(job, sb.toString());
FileOutputFormat.setOutputPath(job, new Path(outputPathStr));
boolean success = job.waitForCompletion(true);
return success ? 0 : 1;
}
public void readFields(DataInput in) throws IOException {
input.readFields(in);
output = Text.readString(in);
}
/**
* Launch application for the dag represented by this client.
*
* @throws YarnException
* @throws IOException
*/
public void startApplication() throws YarnException, IOException {
Class<?>[] defaultClasses;
if (applicationType.equals(YARN_APPLICATION_TYPE)) {
// TODO restrict the security check to only check if security is enabled for webservices.
if (UserGroupInformation.isSecurityEnabled()) {
defaultClasses = DATATORRENT_SECURITY_CLASSES;
} else {
defaultClasses = DATATORRENT_CLASSES;
}
} else {
throw new IllegalStateException(applicationType + " is not a valid application type.");
}
LinkedHashSet<String> localJarFiles = findJars(dag, defaultClasses);
if (resources != null) {
localJarFiles.addAll(resources);
}
YarnClusterMetrics clusterMetrics = yarnClient.getYarnClusterMetrics();
LOG.info(
"Got Cluster metric info from ASM"
+ ", numNodeManagers="
+ clusterMetrics.getNumNodeManagers());
// GetClusterNodesRequest clusterNodesReq = Records.newRecord(GetClusterNodesRequest.class);
// GetClusterNodesResponse clusterNodesResp =
// rmClient.clientRM.getClusterNodes(clusterNodesReq);
// LOG.info("Got Cluster node info from ASM");
// for (NodeReport node : clusterNodesResp.getNodeReports()) {
// LOG.info("Got node report from ASM for"
// + ", nodeId=" + node.getNodeId()
// + ", nodeAddress" + node.getHttpAddress()
// + ", nodeRackName" + node.getRackName()
// + ", nodeNumContainers" + node.getNumContainers()
// + ", nodeHealthStatus" + node.getHealthReport());
// }
List<QueueUserACLInfo> listAclInfo = yarnClient.getQueueAclsInfo();
for (QueueUserACLInfo aclInfo : listAclInfo) {
for (QueueACL userAcl : aclInfo.getUserAcls()) {
LOG.info(
"User ACL Info for Queue"
+ ", queueName="
+ aclInfo.getQueueName()
+ ", userAcl="
+ userAcl.name());
}
}
// Get a new application id
YarnClientApplication newApp = yarnClient.createApplication();
appId = newApp.getNewApplicationResponse().getApplicationId();
// Dump out information about cluster capability as seen by the resource manager
int maxMem = newApp.getNewApplicationResponse().getMaximumResourceCapability().getMemory();
LOG.info("Max mem capabililty of resources in this cluster " + maxMem);
int amMemory = dag.getMasterMemoryMB();
if (amMemory > maxMem) {
LOG.info(
"AM memory specified above max threshold of cluster. Using max value."
+ ", specified="
+ amMemory
+ ", max="
+ maxMem);
amMemory = maxMem;
}
if (dag.getAttributes().get(LogicalPlan.APPLICATION_ID) == null) {
dag.setAttribute(LogicalPlan.APPLICATION_ID, appId.toString());
}
// Create launch context for app master
LOG.info("Setting up application submission context for ASM");
ApplicationSubmissionContext appContext = Records.newRecord(ApplicationSubmissionContext.class);
// set the application id
appContext.setApplicationId(appId);
// set the application name
appContext.setApplicationName(dag.getValue(LogicalPlan.APPLICATION_NAME));
appContext.setApplicationType(this.applicationType);
if (YARN_APPLICATION_TYPE.equals(this.applicationType)) {
// appContext.setMaxAppAttempts(1); // no retries until Stram is HA
}
// Set up the container launch context for the application master
ContainerLaunchContext amContainer = Records.newRecord(ContainerLaunchContext.class);
// Setup security tokens
// If security is enabled get ResourceManager and NameNode delegation tokens.
// Set these tokens on the container so that they are sent as part of application submission.
// This also sets them up for renewal by ResourceManager. The NameNode delegation rmToken
// is also used by ResourceManager to fetch the jars from HDFS and set them up for the
// application master launch.
if (UserGroupInformation.isSecurityEnabled()) {
Credentials credentials = new Credentials();
String tokenRenewer = conf.get(YarnConfiguration.RM_PRINCIPAL);
if (tokenRenewer == null || tokenRenewer.length() == 0) {
throw new IOException("Can't get Master Kerberos principal for the RM to use as renewer");
}
// For now, only getting tokens for the default file-system.
FileSystem fs = StramClientUtils.newFileSystemInstance(conf);
try {
final Token<?> tokens[] = fs.addDelegationTokens(tokenRenewer, credentials);
if (tokens != null) {
for (Token<?> token : tokens) {
LOG.info("Got dt for " + fs.getUri() + "; " + token);
}
}
} finally {
fs.close();
}
addRMDelegationToken(tokenRenewer, credentials);
DataOutputBuffer dob = new DataOutputBuffer();
credentials.writeTokenStorageToStream(dob);
ByteBuffer fsTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength());
amContainer.setTokens(fsTokens);
}
// set local resources for the application master
// local files or archives as needed
// In this scenario, the jar file for the application master is part of the local resources
Map<String, LocalResource> localResources = new HashMap<String, LocalResource>();
// copy required jar files to dfs, to be localized for containers
FileSystem fs = StramClientUtils.newFileSystemInstance(conf);
try {
Path appsBasePath =
new Path(StramClientUtils.getDTDFSRootDir(fs, conf), StramClientUtils.SUBDIR_APPS);
Path appPath = new Path(appsBasePath, appId.toString());
String libJarsCsv = copyFromLocal(fs, appPath, localJarFiles.toArray(new String[] {}));
LOG.info("libjars: {}", libJarsCsv);
dag.getAttributes().put(LogicalPlan.LIBRARY_JARS, libJarsCsv);
LaunchContainerRunnable.addFilesToLocalResources(
LocalResourceType.FILE, libJarsCsv, localResources, fs);
if (archives != null) {
String[] localFiles = archives.split(",");
String archivesCsv = copyFromLocal(fs, appPath, localFiles);
LOG.info("archives: {}", archivesCsv);
dag.getAttributes().put(LogicalPlan.ARCHIVES, archivesCsv);
LaunchContainerRunnable.addFilesToLocalResources(
LocalResourceType.ARCHIVE, archivesCsv, localResources, fs);
}
if (files != null) {
String[] localFiles = files.split(",");
String filesCsv = copyFromLocal(fs, appPath, localFiles);
LOG.info("files: {}", filesCsv);
dag.getAttributes().put(LogicalPlan.FILES, filesCsv);
LaunchContainerRunnable.addFilesToLocalResources(
LocalResourceType.FILE, filesCsv, localResources, fs);
}
dag.getAttributes().put(LogicalPlan.APPLICATION_PATH, appPath.toString());
if (dag.getAttributes().get(OperatorContext.STORAGE_AGENT) == null) {
/* which would be the most likely case */
Path checkpointPath = new Path(appPath, LogicalPlan.SUBDIR_CHECKPOINTS);
// use conf client side to pickup any proxy settings from dt-site.xml
dag.setAttribute(
OperatorContext.STORAGE_AGENT, new FSStorageAgent(checkpointPath.toString(), conf));
}
if (dag.getAttributes().get(LogicalPlan.CONTAINER_OPTS_CONFIGURATOR) == null) {
dag.setAttribute(
LogicalPlan.CONTAINER_OPTS_CONFIGURATOR, new BasicContainerOptConfigurator());
}
// Set the log4j properties if needed
if (!log4jPropFile.isEmpty()) {
Path log4jSrc = new Path(log4jPropFile);
Path log4jDst = new Path(appPath, "log4j.props");
fs.copyFromLocalFile(false, true, log4jSrc, log4jDst);
FileStatus log4jFileStatus = fs.getFileStatus(log4jDst);
LocalResource log4jRsrc = Records.newRecord(LocalResource.class);
log4jRsrc.setType(LocalResourceType.FILE);
log4jRsrc.setVisibility(LocalResourceVisibility.APPLICATION);
log4jRsrc.setResource(ConverterUtils.getYarnUrlFromURI(log4jDst.toUri()));
log4jRsrc.setTimestamp(log4jFileStatus.getModificationTime());
log4jRsrc.setSize(log4jFileStatus.getLen());
localResources.put("log4j.properties", log4jRsrc);
}
if (originalAppId != null) {
Path origAppPath = new Path(appsBasePath, this.originalAppId);
LOG.info("Restart from {}", origAppPath);
copyInitialState(origAppPath);
}
// push logical plan to DFS location
Path cfgDst = new Path(appPath, LogicalPlan.SER_FILE_NAME);
FSDataOutputStream outStream = fs.create(cfgDst, true);
LogicalPlan.write(this.dag, outStream);
outStream.close();
Path launchConfigDst = new Path(appPath, LogicalPlan.LAUNCH_CONFIG_FILE_NAME);
outStream = fs.create(launchConfigDst, true);
conf.writeXml(outStream);
outStream.close();
FileStatus topologyFileStatus = fs.getFileStatus(cfgDst);
LocalResource topologyRsrc = Records.newRecord(LocalResource.class);
topologyRsrc.setType(LocalResourceType.FILE);
topologyRsrc.setVisibility(LocalResourceVisibility.APPLICATION);
topologyRsrc.setResource(ConverterUtils.getYarnUrlFromURI(cfgDst.toUri()));
topologyRsrc.setTimestamp(topologyFileStatus.getModificationTime());
topologyRsrc.setSize(topologyFileStatus.getLen());
localResources.put(LogicalPlan.SER_FILE_NAME, topologyRsrc);
// Set local resource info into app master container launch context
amContainer.setLocalResources(localResources);
// Set the necessary security tokens as needed
// amContainer.setContainerTokens(containerToken);
// Set the env variables to be setup in the env where the application master will be run
LOG.info("Set the environment for the application master");
Map<String, String> env = new HashMap<String, String>();
// Add application jar(s) location to classpath
// At some point we should not be required to add
// the hadoop specific classpaths to the env.
// It should be provided out of the box.
// For now setting all required classpaths including
// the classpath to "." for the application jar(s)
// including ${CLASSPATH} will duplicate the class path in app master, removing it for now
// StringBuilder classPathEnv = new StringBuilder("${CLASSPATH}:./*");
StringBuilder classPathEnv = new StringBuilder("./*");
String classpath = conf.get(YarnConfiguration.YARN_APPLICATION_CLASSPATH);
for (String c :
StringUtils.isBlank(classpath)
? YarnConfiguration.DEFAULT_YARN_APPLICATION_CLASSPATH
: classpath.split(",")) {
if (c.equals("$HADOOP_CLIENT_CONF_DIR")) {
// SPOI-2501
continue;
}
classPathEnv.append(':');
classPathEnv.append(c.trim());
}
env.put("CLASSPATH", classPathEnv.toString());
// propagate to replace node managers user name (effective in non-secure mode)
env.put("HADOOP_USER_NAME", UserGroupInformation.getLoginUser().getUserName());
amContainer.setEnvironment(env);
// Set the necessary command to execute the application master
ArrayList<CharSequence> vargs = new ArrayList<CharSequence>(30);
// Set java executable command
LOG.info("Setting up app master command");
vargs.add(javaCmd);
if (dag.isDebug()) {
vargs.add("-agentlib:jdwp=transport=dt_socket,server=y,suspend=n");
}
// Set Xmx based on am memory size
// default heap size 75% of total memory
vargs.add("-Xmx" + (amMemory * 3 / 4) + "m");
vargs.add("-XX:+HeapDumpOnOutOfMemoryError");
vargs.add("-XX:HeapDumpPath=/tmp/dt-heap-" + appId.getId() + ".bin");
vargs.add("-Dhadoop.root.logger=" + (dag.isDebug() ? "DEBUG" : "INFO") + ",RFA");
vargs.add("-Dhadoop.log.dir=" + ApplicationConstants.LOG_DIR_EXPANSION_VAR);
vargs.add(String.format("-D%s=%s", StreamingContainer.PROP_APP_PATH, dag.assertAppPath()));
if (dag.isDebug()) {
vargs.add("-Dlog4j.debug=true");
}
String loggersLevel = conf.get(DTLoggerFactory.DT_LOGGERS_LEVEL);
if (loggersLevel != null) {
vargs.add(String.format("-D%s=%s", DTLoggerFactory.DT_LOGGERS_LEVEL, loggersLevel));
}
vargs.add(StreamingAppMaster.class.getName());
vargs.add("1>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/AppMaster.stdout");
vargs.add("2>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/AppMaster.stderr");
// Get final command
StringBuilder command = new StringBuilder(9 * vargs.size());
for (CharSequence str : vargs) {
command.append(str).append(" ");
}
LOG.info("Completed setting up app master command " + command.toString());
List<String> commands = new ArrayList<String>();
commands.add(command.toString());
amContainer.setCommands(commands);
// Set up resource type requirements
// For now, only memory is supported so we set memory requirements
Resource capability = Records.newRecord(Resource.class);
capability.setMemory(amMemory);
appContext.setResource(capability);
// Service data is a binary blob that can be passed to the application
// Not needed in this scenario
// amContainer.setServiceData(serviceData);
appContext.setAMContainerSpec(amContainer);
// Set the priority for the application master
Priority pri = Records.newRecord(Priority.class);
pri.setPriority(amPriority);
appContext.setPriority(pri);
// Set the queue to which this application is to be submitted in the RM
appContext.setQueue(queueName);
// Submit the application to the applications manager
// SubmitApplicationResponse submitResp = rmClient.submitApplication(appRequest);
// Ignore the response as either a valid response object is returned on success
// or an exception thrown to denote some form of a failure
String specStr =
Objects.toStringHelper("Submitting application: ")
.add("name", appContext.getApplicationName())
.add("queue", appContext.getQueue())
.add("user", UserGroupInformation.getLoginUser())
.add("resource", appContext.getResource())
.toString();
LOG.info(specStr);
if (dag.isDebug()) {
// LOG.info("Full submission context: " + appContext);
}
yarnClient.submitApplication(appContext);
} finally {
fs.close();
}
}
public void write(DataOutput out) throws IOException {
input.write(out);
Text.writeString(out, output);
}
